A data-driven approach for predicting the ballistic resistance of elastoplastic materials

Data-driven methods and machine learning methods provide efficient and accurate approaches for solving impact problems. In this paper, a data-driven approach is proposed for numerical simulations of ballistic impact behavior for elastoplastic materials. An enhanced rate-dependent scheme is employed for improving the predictability of the data-driven constitutive model. A new method that introduces a stress triaxiality indicator to three separate constitutive models is proposed to consider the discrepancy between the mechanical responses of materials under tension, compression, and shear. Additionally, a modified Bai-Wierzbicki fracture criterion considering the strain rate effect and the stress-state effect is used to evaluate the fracture behavior of the materials during impact simulations. Subsequently, a compatible numerical implementation algorithm that considers loading, unloading, and reverse loading is established to enable the application of the data-driven approach in finite element simulations. Numerical validation of the proposed data-driven approach is conducted through several simple loading examples, such as cyclic loading, torsional loading, and tension-torsion combined loading. The data-driven approach is then employed to simulate the ballistic impact behavior of Ti-6Al-4V targets of different thicknesses that are struck by blunt projectiles. Impact properties-including the relationship between residual velocity and impact velocity, ballistic limit velocities, and fracture paths-are comprehensively studied. The results demonstrate the reasonable predictability and accuracy of the data-driven approach when applied to ballistic impact simulations.